Atlanta Asks Google Whether It Targeted Black Homeless People
nytimes.com
nytimes.com
As far as I can see, their options are:
- Keep using datasets largely made up of white college kids and stock photo models, and be accused of racism because their image recognition systems don't work so well for other demographics
- Specifically seek out other demographics to broaden their datasets, and be accused of racism for it
- ...stop trying to make any kind of system that deals with images of humans?
It is simply a false dichotomy that their data samples are either white college kids or whoever they seek offline as the presumed opposite.
It is more work to get a reasonable dataset instead of a bad one and one or two patches. I think they can afford that work using their normal portal in a way that isn't particularly controvesial aside from the public debates that should occur in general.
That's literally dehumanizing.
Speaking specifically of my family and some colleagues from western Europe, they just don't seem to have an appreciation for how high strung equality and race relations are in America and regularly say shit that makes me wince (and laugh, usually). Having a team from the Netherlands run this, I'm completely unsurprised that this rather tone deaf but profoundly utilitarian and effective approach is where they landed.
An American company may or may not have picked an ethically superior approach but they almost certainly would have picked an optically superior one.
It's a multinational company with branches everywhere. It's only Dutch by its origins and HQ.